Assignments
Prof. Jesús Labarta
BSC & UPC
Agenda
• Introduction: before start working with the system
• Software requirements, login information, configuration, system overview, etc.
• Paraver tutorial: getting familiar with paraver
• A Paraver tutorial to introduce the fundamentals of this tool
• The matrix multiply example: playing with this code
• The matrix multiplication code has appeared during the seminar’s slides. A good opportunity to try it!
• Tutorial exercises: applying learned lessons (guided examples)
• Heat and N-Body kernels using TAMPI
• LULESH using DLB
• The IFSKer Practice: let's do it!
• Submit your report about your experiences applying TAMPI or DLB
Documents & slides (available on-line)
https://pm.bsc.es/ftp/training/acm-ess-2019
4
Software requirements
• SSH: Secure SHell (to connect the HPC system)
• Linux: has native support of secure shell “ssh user@host”
• Windows: need to install a ssh program
• PuTTY http://www.chiark.greenend.org.uk/~sgtatham/putty/download.html
• MobaXterm http://mobaxterm.mobatek.net/download.html
• X Server / X forwarding (for wxparaver or .pdf readers)
• Linux: has native support (remember to connect with “ssh -X user@host”
• Windows: need to install a X server program
• Xming https://sourceforge.net/projects/xming/
• MobaXterm already includes a X server within the package
Login information
• Connecting to Nord III (login)
• nord1.bsc.es
• nord2.bsc.es
• User provided in your badge
John
Smith
Student
nct01XXX
Practical information for the ACM Summer School
Xxxxx xxxx xx xxxxx xxxx x xxxxxxxx xx xxxxx xxxx x xxxx xxxxx xxxx xxxxxx xxxx xxxxxx x xx xxxxxx xxxx xx xxxxx xxxxxxxx xx xxxxx xxxx x xxxx
xxxxx xxxx xxxxxx x xx xxxxxx x xxxx xxxxx xxxx xxxxxx x xx xxxxxx xxxx x xxxx xxxxx x xxxx xxxxx xxxx xxxxxx x xx xxxxxx xxxx xxxxxx xxxxxxxxx x xx xxxxxx
xx xxxxx xxxx xx xxxxx xxxx x xxxx xxxxx xxxx xxxxxx x xx xxxxxx xxxx x xxxx xxxxx xxxx xxxxxx x xx xxxxxx
WIFI in Sala Àgora Xxxxxx
Xxxxxx xxxxxxxx
Access to FTP Server for lecture material xxxxx xxxx xxxxxx x xx xxxxxx xxxx
Xxx xxxxxx x xxxx Xxx xx xxxxx x xxxx
Access to Nord cluster ssh [email protected]
Username: nct01YYY (specific value for YYY included in badge) Password: ●●●●●●●●●●
6
PuTTY: SSH configuration
Before open session make sure…
MobaXterm: SSH configuration
8
System Overview: Nord III
• 9 iDataPlex compute racks. Each one composed of:
• 84 IBM dx360 M4 compute nodes
• 4 Mellanox 36-port Managed FDR10 IB Switches
• 2 BNT RackSwitch G8052F (Management Network)
• 2 BNT RackSwitch G8052F (GPFS Network)
• 4 Power Distribution Units
• All IBM dx360 M4 node contain:
• 2x E5–2670 SandyBridge-EP 2.6GHz cache 20MB 8-core
• 500GB 7200 rpm SATA II local HDD
• 8x 16G DDR3–1600 DIMMs (8GB/core) Total RAM: 128GB/node
• 1.9 PB of GPFS disk storage
• Interconnection Networks
• Infiniband Mellanox FDR10: High bandwidth network used by parallel applications communications (MPI)
• Gigabit Ethernet: 10GbitEthernet network used by the GPFS Filesystem.
• Operating System: Linux - SuSe Distribution 11 SP3
Assignment description
https://pm.bsc.es/ftp/training/acm-ess-2019
10
01 – Paraver Tutorial
• Getting familiar with Paraver
• You can overlap this exercise with the next matrix multiplication example
• Get traces, analyze them, get the corresponding histograms, etc.
• Complete this sections with the Extrae/Paraver documentation [see more]:
https://tools.bsc.es/tools_manuals
02 – Matrix Multiplication example
• This code has been used through this tutorial
• Several version/techniques are included in the slides
• Overlap computation in tasks and communication in master
• Overlap between computation and communication in tasks (+ nested)
• Can start computation of next block of C as soon as communication terminates
• Can obtain parallelism without
parallelizing fine grain the computation
• Now, you can try it, change it, test it
12
03 – TAMPI Exercises
• N-Body simulation kernel: system of bodies interacting one each other
• Heat diffusion kernel: Gauss-Seidel method solving the heat equation
• Goals of this exercise
• Analyze execution, is there comm.
serialization?
• Adapt the code to exploit TAMPI
• Compare performance results with the initial version
• Compare traces before/after changes
04 – DLB Exercises
• LULESH: The Shock Hydrodynamics Challenge Problem
• Goals of this exercise
• Analyze execution, is there imbalance?
• Adapt program in order to use DLB
• Compare performance results with the initial version
• Compare traces before/after changes
14
05 – IFSKER Practice
• IFSKernel code: a kernel of the OpenIFS application, used in global forecast and climate reanalyzes (from ECMWF)
• Consider using other interesting codes:
• NTCHEM, Your own application! Send us an email1 to confirm “interest”
• Goals of this exercise
• Analysis Phase: check program’s arguments and/or runtime’s options impacting on the behavior (e.g., number of threads, schedulers, etc.). Scalability/parallel
efficiency. Traces. Histograms.
• Argument: Is there imbalance? Is there comm. serialization? Others?
• Determine (and justify) what is your approach:
• Consider to adapt the code to exploit DLB or TAMPI
• Is there any other technique you consider it can improve the code?
• Compare performance results and traces before/after your contribution
• Highlights. Conclusions. Extensions/next-steps.
______________________
1 Send an email to htbp-submit [at] bsc.es
05 – IFSKER Practice (submit)
• Write a 5-10 pages document report. It should include:
• Initial performance analysis (how the application behaves)
• Scaling/efficiency factors according to…
• Paraver traces/histograms
• Others,…
• Proof of concept description (your plan)
• Implementation description (what you have done)
• Parallelization, taskyfication, synchronization, etc.
• Experiment results (how the application behaves now)
• Version comparison, discussion, compare them with initial analysis
• Experiences and conclusions (what you have learnt)
